research open access genomic occupancy of runx2 with ... · research open access genomic occupancy...

Wu et al. Genome Biology 2014, 15:R52http://genomebiology.com/2014/15/3/R52

RESEARCH Open Access

Genomic occupancy of Runx2 with globalexpression profiling identifies a novel dimensionto control of osteoblastogenesisHai Wu1, Troy W Whitfield2, Jonathan A R Gordon1, Jason R Dobson3,4, Phillip W L Tai1, Andre J van Wijnen5,Janet L Stein1, Gary S Stein1 and Jane B Lian1*

Abstract

Background: Osteogenesis is a highly regulated developmental process and continues during the turnover andrepair of mature bone. Runx2, the master regulator of osteoblastogenesis, directs a transcriptional program essentialfor bone formation through genetic and epigenetic mechanisms. While individual Runx2 gene targets have beenidentified, further insights into the broad spectrum of Runx2 functions required for osteogenesis are needed.

Results: By performing genome-wide characterization of Runx2 binding at the three major stages of osteoblastdifferentiation - proliferation, matrix deposition and mineralization - we identify Runx2-dependent regulatorynetworks driving bone formation. Using chromatin immunoprecipitation followed by high-throughput sequencingover the course of these stages, we identify approximately 80,000 significantly enriched regions of Runx2 bindingthroughout the mouse genome. These binding events exhibit distinct patterns during osteogenesis, and are associatedwith proximal promoters and also non-promoter regions: upstream, introns, exons, transcription termination site regions,and intergenic regions. These peaks were partitioned into clusters that are associated with genes in complex biologicalprocesses that support bone formation. Using Affymetrix expression profiling of differentiating osteoblasts depleted ofRunx2, we identify novel Runx2 targets including Ezh2, a critical epigenetic regulator; Crabp2, a retinoic acid signalingcomponent; Adamts4 and Tnfrsf19, two remodelers of the extracellular matrix. We demonstrate by luciferase assays thatthese novel biological targets are regulated by Runx2 occupancy at non-promoter regions.

Conclusions: Our data establish that Runx2 interactions with chromatin across the genome reveal novel genes, pathwaysand transcriptional mechanisms that contribute to the regulation of osteoblastogenesis.

BackgroundThe development of bone tissue, new bone formation inthe adult, mineral homeostasis and maintenance of bonemass are mediated by cells of the osteoblast lineage. Thesebone forming cells progress through a highly regulated dif-ferentiation program, with each subpopulation of cells ac-quiring stage-specific phenotypes that are characterized bydistinct profiles of expressed genes [1]. Osteoblast com-mitment and differentiation are dependent on the appro-priate expression of Runx2 (Runt-related transcriptionfactor 2), the master regulator of bone formation [1-3].Ablation of Runx2 in mice results in the absence of a

* Correspondence: [email protected] of Biochemistry, University of Vermont College of Medicine andVermont Cancer Center, 89 Beaumont Avenue, Burlington, VT 05405, USAFull list of author information is available at the end of the article

© 2014 Wu et al.; licensee BioMed Central LtdCommons Attribution License (http://creativecreproduction in any medium, provided the or

mineralized skeleton, and disruption of Runx2 functioncauses bone defects in the human disorder cleidocranialdysplasia [4,5]. Runx2 controls a complex gene-regulatorynetwork during osteoblastogenesis [5,6]. It upregulates avariety of osteoblast lineage-specific genes, including Osx(osterix), Ocn (osteocalcin), and Bsp (bone sialoprotein),and represses the expression of non-osteoblast genes suchas PPAR-γ (peroxisome proliferator-activated receptorgamma) and MyoD (myogenic differentiation), which arerequired for adipogenic and myogenic commitment, re-spectively [7-9]. Studies have demonstrated that Runx2controls gene expression by interacting with multiple clas-ses of co-regulatory factors [1,3,10]. In addition to thetraditional transcriptional mechanisms, several epigeneticmechanisms have been identified for Runx2-mediatedgene expression. For example, Runx2 is retained on

. This is an Open Access article distributed under the terms of the Creativeommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andiginal work is properly credited.

mailto:[email protected]

http://creativecommons.org/licenses/by/2.0

Day 0 9 21 28

ALP

Von

Kos

sa

A

B

BspAkp2

Ocn

21 280 9 day

10000

1000

0

100

10

1

Osx

0

2

6

8

10

4

0 9 21 28day

Col1a1re

lativ

e m

RN

A e

xpre

ssio

n

C

0 9 21 28Day 14

Runx2

H3

0

2

6

8

10

4

0 9 21 28day

Runx2

rela

tive

mR

NA

leve

l

Figure 1 The differentiation stages of murine MC3T3-E1preosteoblasts used for profiling studies. (A) Staining for alkalinephosphatase (ALP) activity (upper panel) and mineralization (Von Kossa,lower panel) in MC3T3-E1 cells during three stages of differentiation:proliferation (day 0), matrix deposition (days 9 to 21), and mineralization(day 28). (B) Protein (left panel) and mRNA levels (right panel) of Runx2during osteoblastic differentiation of MC3T3-E1 cells. Histone H3 (H3) wasused as loading control for western blotting. (C) Expression profile ofosteoblast-related markers Osx/Sp7, Col1a1 (left panel), Akp2/Alpl (alkalinephosphatase liver/bone/kidney), Bsp/Ibsp, and Ocn/Bglap2 (right panel) inMC3T3-E1 cells during differentiation. Relative mRNA levels (versus day 0)were determined by quantitative RT-PCR (reverse transcription PCR),normalized by Hprt1 mRNA levels and plotted as mean values ± SEM(standard error of mean) from three independent biological replicates.The expression levels of the genes in (B,C) at days 9, 21, and 28 weresignificantly upregulated when compared to those at day 0 (P<0.05, t-test).

Wu et al. Genome Biology 2014, 15:R52 of 17http://genomebiology.com/2014/15/3/R52

mitotic chromosomes as an epigenetic bookmarking factorto maintain cellular identity after cell division [11]. A sig-nificant body of evidence substantiates the contribution ofRunx2 towards the epigenetic regulation of gene expres-sion during osteoblast differentiation, through interactionswith histone deacetylases [12], histone acetyltransferases[13], and SWI/SNF complex components [14].Runx2 has been shown to activate or repress gene expres-

sion through binding to cis-regulatory DNA elements, theRunx motif (TGTGGT) located in or near gene promoterregions [15]. Genome-wide binding profiles of the tran-scriptional factor CTCF and lineage-specific transcriptionfactors, such as PPAR-γ, MyoD, and GATA3 (GATA bind-ing protein 3), via ChIP-Seq (chromatin immunoprecipita-tion followed by high-throughput sequencing) have helpedto delineate cis-regulatory networks that are critical for celllineage control in adipocytes, myocytes and T cells, respect-ively [16-19]. These studies have highlighted the regulatoryimportance of long-range interactions and binding ofphenotypic transcription factors to non-promoter genomicelements. These findings have greatly expanded existingparadigms of transcriptional regulation. In contrast, the en-tire scope of Runx2 binding elements remains largely un-known, hindering a comprehensive understanding of thecis-regulatory network through which Runx2 regulates thetranscription program for bone formation.Hence, we have characterized the genome-wide occu-

pancy of Runx2 by ChIP-Seq in MC3T3-E1 preosteoblasts,a well-studied in vitro model for osteoblastogenesis [20].Runx2 occupancy was determined at three hallmark stagesof osteoblast differentiation: proliferation, matrix depos-ition, and mineralization. Analyses of these data demon-strated that Runx2 occupancy on genes is differentiationstage-dependent in both binding intensities and bindingregions, indicating a shift in the regulatory mechanismsrequired for the entire program of osteoblastogenesis. Bycoupling genome-wide Runx2 binding with gene expres-sion profiling, we have identified new Runx2 targets thatwere validated for functional activities of Runx2 binding inboth promoter and non-promoter regions. Our study ofRunx2 genome-wide occupancy establishes a foundationfor future investigation of the Runx2-controlled regulatorynetwork during bone formation and homeostasis.

ResultsRunx2 dynamically occupies a wide range of genomic lociduring osteoblastogenesisRunx2 is a known master activator of bone formation,but thus far only a small number of osteoblast-specifictarget genes have been characterized [2,5,6]. To identifygenome-wide occupancy of Runx2 in osteoblast lineagecells, we performed ChIP-Seq using a Runx2-specificantibody at stages during the in vitro differentiation ofMC3T3-E1 preosteoblasts (Figure 1). This in vitro cell

model recapitulates in vivo osteoblast differentiation andtherefore was used for ChIP-Seq studies [21]. Alkalinephosphatase activity, an early osteoblast marker, in-creased as cells proceeded from proliferation (day 0) tomatrix deposition (days 9 and 21) and decreased uponmineralization (day 28), visualized by Von Kossa staining(Figure 1A). Runx2 mRNA and protein levels signifi-cantly increased during the initial stage of differenti-ation. While the mRNA levels reached steady state,Runx2 protein levels declined during late mineralization(Figure 1B). Osteoblast phenotypic markers, includingthe transcription factor Osx/Sp7 (osterix/Sp7 transcrip-tion factor), a marker of committed osteoprogenitors,


the extracellular matrix protein Col1a1 (collagen type Ialpha 1), and the specialized mineral binding proteinsBsp/Ibsp (bone sialoprotein/integrin-binding sialoprotein)and Ocn/Bglap2 (osteocalcin/bone gamma-carboxyglutamate(gla) protein 2), displayed expression patterns consistent withthe progression of osteoblastogenesis [2] (Figure 1C).We next isolated and sequenced DNA from chromatin

bound by Runx2. Sequence reads from Runx2 ChIPs and in-put controls were mapped to the mouse genome. Statisti-cally significant enrichments of Runx2 were identified byMACS (Model-Based Analysis of ChIP-Seq) [22] (Additionalfile 1). RefSeq annotations [23] were used to assign Runx2enrichments to categories of genomic locations (with non-overlapping definitions for transcriptional start site (TSS),promoter, exonic, intronic, transcription termination site(TTS), upstream and intergenic regions (see Figure 2 for

A

0

10

20

30

40

50

60

70

0%

20%

40%

60%

80%

100%

Num

ber

of R

unx2

pea

ks (

x103 )

0 9 28 0 9 28

perc

ent o

f Run

x2 p

eaks

of t

otal

2

1

0

bits

9 10

2

1

0

bits

1 2 3 4 5 6 7 8

1 2 3 4 5 6 7 8 9 10 11

C

Day Day

Figure 2 Genome-wide profile of Runx2 occupancy. (A) Distribution ofsix categories of genomic locations: exon, intron, promoter (-1 kb to +150to -1 kb of transcription termination site), and intergenic region. Peak distriand by the percentage of total peaks (right panel). (B) Differential Runx2 enpredicted Runx2 motif (see below), to random binding, or to binding of theand TTS regions. (C) The 500 most significant Runx2 peaks, based on MACS [22]motif (position weight matrix, top) with strong statistical confidence (P= 4.7 × 10Runx motif (MA0002.2, bottom) in JASPAR [26] was used for comparison usingamong categories of genomic locations was determined using FIMO [24] at aof Runx2 peaks indicating the distances of Runx2 motifs to the peak centers iscores) versus the probabilities of finding de novo Runx2 motifs at given posit

details; Additional file 2). As Runx2 protein levels changed,the total number of Runx2 peaks changed accordingly(Figure 2A, left panel). The overall distribution of Runx2binding among the categories of genomic locations was rela-tively unchanged during differentiation (Figure 2A, rightpanel). Among genomic locations, Runx2 occupancy at pro-moters showed the greatest variation from 17.6% at day 0 to8.8% at day 9 (Figure 2A, right panel). The majority ofRunx2 binding occurred at intergenic and intronic regions(Figure 2A,B). However, when compared with a randomlysampled background distribution of genomic intervals(Additional file 3), Runx2 binding displayed preferential en-richment in a genic context, particularly at promoters andexons. One exception was under-represented Runx2 bind-ing at intergenic regions when compared to nonspecific orrandom binding (Figure 2B). This relative enrichment of

IntergenicTTSPromoterIntronExonUpstream

D

Frac

tion

ofR

unx 2

bind

ing

loci

Inter

genic

Upstre

am

Prom

oter

Intro

nExo

n

0.7

0.6

0.5

0.4

0.3

0.2

0.1

B

TTS0.0

Runx2CTCF

Runx2 motifrandom

Promoter TTS0.00

0.08

0.06

0.04

0.02

Frac

tion

ofR

unx 2

loci

-1000 -500 0 500 1000

Pro

babi

lity

(Dis

tanc

e)

Distance(bp)

0.006

0.005

0.004

0.003

0.002

0.001

0

top 1/3 of peaks

middle 1/3 of peaks

bottom 1/3 of peaks

Runx2 binding peaks across the mouse genome were classified intobp of TSS), upstream (-1 kb to -20 kb from TSS), TTS region (-150 bpbution was plotted at each time point by peak number (left panel)richment in the six categories of genomic locations compared to thetranscription factor CTCF. Inset provides a magnified view of promoter

significance (P< 1× 10-10) were used for de novomotif discovery. A Runx2-200) was determined using MEME (MEME suite version 4.7.0) [24]. The knownTOMTOM [24]. As shown in (B), the distribution of de novo Runx2 motifssignificance threshold of P < 10-4. (D) Probability plot of the distributionn the top, middle, and bottom third of Runx2 peaks (ranked by MACSions relative to peak center.


Runx2 occupancy in genic contexts was also observed incomparison with Runx2 motifs (Figure 2B). A de novoRunx2 motif (Figure 2C, top) was discovered using MEME[24] on the highest-ranked (by P-value) 500 ChIP-Seqpeaks and was then scanned using FIMO [24] (version4.7.0) over the mouse genome. Notably, analysis by MEMEdid not discover a significantly enriched secondary motifassociated with Runx2 peaks. Despite the nearly randomdistribution of Runx2 motifs throughout the genome,Runx2 occupancy in differentiating osteoblasts is character-ized by associations with promoters, exons, introns andother genic elements. These associations are perhaps dueto epigenetic factors, including chromatin conformationand accessibility, along with co-factors, and suggest thatthe presence of a Runx motif does not necessarily indicatethe physical association of Runx2. Interestingly, the distri-bution of Runx2 occupancy among classes of genomic ele-ments is similar to that of CTCF (Figure 2B), a ubiquitoustranscription factor that exhibits a broad spectrum of DNAbinding in many cell lines [25].The de novo Runx2 motif (Figure 2C, top) was scanned

over all ChIP-Seq peaks called by MACS [22] from ChIP-Seq reads collected from day 9 MC3T3 cells, and a histo-gram of the distance between the peak summit and thehighest-scoring motif instance was collected. The distribu-tion in Figure 2D is characterized by a sharp peak forRunx2 motifs at the summits of ChIP-Seq peaks. UsingTOMTOM [24] it was determined that the de novo Runx2motif was similar (E-value = 2.6 × 10-4, q-value = 2.6 × 10-4)to the JASPAR MA0002.2 RUNX motif [26] (Figure 2C,bottom) and both motifs share the core TGTGGT se-quence with the known Runx2 binding consensus.

Genome-wide Runx2 occupancy reveals distinct positionaland temporal binding patternsWe grouped Runx2 peaks based on the presence or ab-sence of Runx2 binding at specific time points duringdifferentiation (Figure 3A, off/on). The resulting sevendistinct clusters reflected the dynamics of Runx2 bind-ing in relation to the progression of osteoblastogenesis(Figure 3A). In Figure 3B, the mean ChIP-Seq read dens-ities (that is, peak intensities) of the clustered Runx2 peakswere plotted to compare their relative enrichments.The seven clusters differed in the numbers and inten-

sities of Runx2 peaks (Figure 3A). The largest group wascluster 6, which exhibited the presence of peaks primarilyat day 9. This finding is consistent with day 9 representingcommitted osteoblasts with the highest amount of cellularRunx2 protein (Figure 1B) and therefore the greatest num-ber of Runx2 peaks (Figure 3B). The strongest Runx2 peakintensities were found in cluster 1 reflecting regions thatwere bound by Runx2 constitutively throughout osteoblastdifferentiation (Figure 3B). The weakest Runx2 peak inten-sities among the three time points occurred at the peaks

in cluster 7, with marginal enrichment of Runx2 at day 0.It is noteworthy that the peaks in cluster 4 have the sec-ond highest peak intensities at both days 9 and 28 (matrixdeposition and mineralization stages; Figure 3B). Cluster5, like clusters 1 and 4, exhibited the highest peak inten-sities on day 28, indicating their importance in maintain-ing osteoblast phenotype.Runx2 binding in each cluster was further examined for

distribution preferences of peaks in different genomic re-gions, in contrast to the genome-wide distribution of ran-dom 100 bp DNA fragments (detailed in Additional file 3;Figure 3C). The random DNA fragments (grey bars inFigure 3C) are distributed mainly in intergenic, intronic,and upstream regions. When compared to the random con-trol, we observed that the distribution of Runx2-enrichedpeaks was biased towards gene regions (exons, introns, pro-moters, TTS regions, upstream regions). In contrast, Runx2binding in the intergenic regions is lower than the ran-dom control (Figure 3C). In promoters and exons, allclusters showed the highest enrichment over randombinding, suggesting strong regulation by Runx2 at thesegenomic regions.To further explore the relationship between Runx2

peaks and peak-associated genes in osteogenic differenti-ation, we performed functional annotations for the peaksin the seven clusters using GREAT (Genomic Regions En-richment of Annotations Tool; Figure 3D). Gene Ontology(GO) terms associated with the largest clusters are shownin Figure 3D. Runx2 binding in cluster 1 yielded GO termsof general biological processes such as protein folding andRNA metabolism (detailed in Additional file 4). Cluster 4peaks at days 9 and 28 were frequently related to the GOterms of osteoblast differentiation, bone developmental pro-cesses, and osteogenic signaling pathways. These terms oftenincluded differentiation-related and well-known Runx2 tar-get genes, such as Runx2, Bsp/Ibsp, and Osx/Sp7 (Additionalfile 4). Similarly, cluster 6 (day 9) peaks often associated withbone formation and extracellular matrix organization. Theannotation of other smaller clusters is shown in Additionalfile 5. For examples, cluster 3 peaks were associated withapoptosis, programmed cell death, and DNA damage; clus-ter 7 containing peaks found only in day 0 was linked withnegative regulation of cell cycle control and the phenotypesof non-osseous mesenchyme-derived cells (Additional files 4and 5). These functional annotations associated with Runx2peaks are generally consistent with the progression ofMC3T3-E1 differentiation, supporting a temporal transcrip-tion network programmed by Runx2.

Runx2 binding patterns at osteogenic genes arepredictive of potential Runx2 targetsTo discover previously unknown Runx2 target genes, wefirst determined the Runx2 binding patterns of well-knownRunx2 target genes found in differentiation cluster 4; for

A

2 4 60

0 9 28Day

0 9 28

off on

1

23

4

5

6

7

Clu

ster

Day CB

TTS

Exon

Fraction of Runx2 binding regions0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Intron

Intergenic

Promoter cluster 1

cluster 4cluster 3cluster 2

cluster 5cluster 6cluster 7random

Upstream

D

mea

n lo

g 2(C

hIP

-Seq

rea

d de

nsit y

+ 1

)

4.0

0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

9 280

Day

cluster 1



protein foldingneg. reg. mitotic cell cycleprotein peptidyl-prolyl isomerizationnuc.-transcribed mRNA catabolic processpeptidyl-proline modificationmRNA catabolic processRNA catabolic processcortical cytoskeleton organizationinduction of apoptosis by intracellular sig.maintenance of protein locationreg. insulin receptor signaling pathway

ubiquitousCluster 1

202

Obs.peaks

10060

12871

135154

7411513284

Obs.genes

114/15222/2827/3638/5031/4143/5855/7619/2238/4952/6819/23

44.94.8scores (-log10(p-value))

day 9

collagen fibril organizationforegut morphogenesishair follicle morphogenesisreg. insulin receptor signaling pathwa yRho protein signal transductionbone remodelingbone resorptionestrogen receptor signaling pathway

Cluster 6Cluster 4 day 9 & 28

cell junction organizationepidermis morphogenesisosteoblast differentiationstem cell developmentosteoblast developmenthair follicle morphogenesistissue remodelingcell junction assemblyembryonic pattern specificationcell-cell junction organizationbone remodeling

252

Obs.peaks

196323288125148227187218170104

Obs.genes54/6630/3660/7046/6315/1821/2658/7937/4640/5238/4623/32

241

Obs.peaks

15521514917016210379

Obs.genes26/2712/1224/2621/2328/3230/3217/1812/12

Figure 3 Characterization of the temporal patterns of Runx2 binding associated with osteoblastogenesis. (A) Enriched regions of Runx2binding (MACS peaks, P ≤ 1 × 10-10) were grouped into seven clusters based on the presence (red) or absence (blue) of peaks at temporal stagesof osteoblastic differentiation (left panel). In the right panel, each line illustrates intensity (base-2 logarithm of the sum of read counts per 10million reads (mean read density)) of one peak as a range from strong occupancy (brown) to weak occupancy (violet). (B) Runx2 bindingintensities of Runx2 peaks from each cluster at days 0, 9, and 28 of differentiation plotted as mean ± SEM. (C) Distribution patterns of the Runx2peaks from each of the seven clusters (from (A)) segregated into six categories of genomic locations and compared to an equivalent number ofrandom 100 bp regions (gray bar). (D) Representative Gene Ontology term annotation of cluster 4 (days 9 & 28) by GREAT. Each term is annotatedwith the observed (Obs.) number of Runx2 peaks and corresponding number of genes out of the total number of genes.


example, Runx2, Osx/Sp7, and Ocn/Bglap2 (Figure 4A;Additional file 6), and Bsp/Ibsp (Additional file 7). For thesegenes Runx2 binding was distributed in promoters, the genebody (introns/exons) and sites distal from the gene body.Furthermore, Runx2 enrichment increased in the loci ofthese genes during osteoblast differentiation at matrix andmineralization stages. We then examined the genes associ-ated with cluster 1 peaks (Additional file 8), and identifiedgenes, including Ezh2 (enchancer of zeste homolog 2)(Figure 4B), with Runx2 binding profiles that displayed aubiquitous but increasing level of Runx2 occupancy. Ezh2 isa component of PRC2 that epigenetically regulates gene ex-pression by methylating histone H3 lysine 27 and wasrecently found to be involved in commitment of mesenchy-mal stem cells towards the osteoblast lineage [27]. The 5′proximal promoter region of Ezh2 is bound by Runx2(Figure 4B; Additional file 6) and shows an increase in reads(occupancy) during differentiation. When Ezh2 mRNA

levels were measured during osteoblast differentiation, wefound that the highest expression occurred in proliferatingMC3T3-E1 cells, when the lowest amount of Runx2 bindingwas observed (Figure 4C). The striking decrease in Ezh2mRNA levels with increased Runx2 binding suggests thatthe transcription of Ezh2 is potentially regulated by Runx2.

Genes regulated by Runx2 exhibit distinct Runx2 bindingprofilesTo characterize the Runx2 binding profiles in genes thatare transcriptionally regulated by Runx2, we performedgene expression profiling at day 9 in MC3T3-E1 cellstreated with control (Scr) short hairpin RNA (shRNA) ora Runx2-specific shRNA (shRunx2) that knocks downRunx2 expression. Runx2 protein levels decreased by 80%in cells treated with shRunx2 (Additional file 9). Thisknockdown in turn inhibited expression of differentiationmarker genes and osteoblastogenesis as demonstrated by

Runx2 Consensus

15

0 15

0 15

0

_

_

_

_

_

_

5 kb

Chr15: 102,185,000 102,199,000

Runx2 Consensus

55

0 55

0 55

0

_

_

_

_

_

_

Osx/Sp7

Chr6:

10

0 10

0 10

0

_

_

_

_

_

_

0

0.2

0.4

0.6

0.8

1

1.2

Day0 Day 9 Day21 Day 28

Ezh2

Rel

ativ

e m

RN

A e

xpre

ssio

n

A100 kb

Chr17: 44,725,000 45,000,000

Runx2 Supt3h

5 kb

Chr3: 88,174,000 88,190,000

Bglap-rs1 Bglap

Runx2 Consensus

55

0

55

0 55

0

_

_

_

_

_

_

Ocn/Bglap2

CB47,480,000 47,550,000Ezh2

Runx2 Consensus

25 kb

day 0

day 9

day 28

input_

_

15

0

input_

_

55

0

input_

_

55

0

day 0

day 9

day 28

day 0

day 9

day 28

day 0

day 9

day 28

10

0

_

_input

Figure 4 Runx2 binding at known osteogenic genes and a potential target, Ezh2, during osteoblastogenesis. (A) Runx2 binding at threeknown osteogenic genes, Runx2, Osx, and Ocn, during three stages of osteogenesis: proliferation (day 0); matrix deposition (day 9); and mineralization(day 28). Gene annotation follows standard gene prediction display conventions used by the UCSC genome browser (exons, solid boxes; introns, solidlines; direction of gene transcription, arrows). Positions of Runx2 peaks called by MACS (green bars) and Runx2 consensus motif (TGTGGT; solid blackbar) are also depicted. (B) Ezh2 locus bound by Runx2 with increasing Runx2 enrichment over temporal stages of osteogenesis. (C) Ezh2 expressionduring osteogenesis (relative to day 0 levels) were determined by quantitative RT-PCR and normalized by Hprt1 mRNA levels. Relative mRNA levels areplotted as mean ± SEM from three independent biological replicates.


decreased Alp staining (Additional file 9). We found 159genes whose expression was responsive to Runx2 knock-down, with |log2(IshRunx2/IScr)| > 1.5, and a false discoveryrate (FDR) <0.05, where I is the measured normalized

probe-set intensity. These genes included 115 up- and 44down-regulated genes (Figure 5A). The 15 genes most re-sponsive to Runx2 knockdown (Table 1) included thewell-defined Runx2 targets Ocn/Bglap2, Bsp/Ibsp, and

BA

log2(Scr expression)16141210864

16

14

12

10

8

6

4

log2

(shR

unx2

exp

ress

ion)

Ly6a

Ly6c2Usp18

Stat2Dkk2 Mmp13

Tnfrsf19Ibsp

AcppBglap2Alpl

Phex

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

Pea

k nu

mbe

r pe

r ge

ne

Peak groups

peak numberDownUp

Non-responsive

All peaks

*

D0 peaks

**

D9 peaks

*

*

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Per

cent

age

of to

tal p

eaks

Intergenic

TTS

Promoter

Intron

Exon

Upstream

*

C

ED

peak distribution*

Down

Non-re

spon

sive Up

0

2

4

6

8

10

12

14

Upstream Exon Intron Promoter TTS Intergenic

D9/

d0 fo

ld c

hang

e of

pea

k si

gnal

s

fold change of peak signal

*

DownUp

Non-responsive

*

*

180

140

100

60

20

Mea

n R

unx2

enr

ichm

ent

at d

ay 9

STTSST-10kb +10kb0

Down

Up

Non-responsive

Gene bodyUpstream DownstreamGenomic location

-1

Mea

n ph

yloP

sco

re0

12

34

Down

Up

Non-responsive

Non-responsive Down

Non-responsive Up

F phyloP score****

**

Figure 5 Correlation of Runx2 occupancy and Runx2-reponsive genes identifies novel targets. (A) Expression levels of genes responsive toRunx2 silencing in differentiated MC3T3-E1 cells. Values are plotted as the log2(expression level) from shRunx2-expressing cells (vertical axis) versus controlshRNA expression (horizontal axis). Each point represents mean mRNA expression level from three independent biological replicates. Several representativegenes are labeled. Diagonal lines demarcate the threshold for significant increase or decrease (≥1.5-fold) in expression. (B-D) Runx2 peaks associated withupregulated (Up), downregulated (Down), or unchanged (Non-responsive) gene expression upon Runx2 knockdown were compared by: average peaknumber per gene (B), and peak distribution (C) and fold change of peak signals (d9 versus d0) across genomic locations (D) with shRunx2 non-responsivegenes as a control. In (B), three groups were compared to non-responsive genes: all peaks in shRunx2-regulated genes (All peaks), all peaks present at day9 in shRunx2-regulated genes (D9 peaks), and all peaks present at day 0 in shRunx2 regulated genes (D0 peaks). Values are mean ± SEM (B,D) and statisticalsignificance (*P< 0.01, **P< 0.05) determined by Mann-Whitney test (B,D) or Fisher’s exact test (C). (E) Runx2 enrichment across gene bodies (±10 kb) ofgenes downregulated (Down), upregulated (Up), and unchanged (Non-responsive) by shRunx2 treatment at day 9. Mean signal ratios (IP/Input) at eachgenomic region were determined using PeaksToGenes. Error bars represent SEM. (F) Mean phyloP conservation scores of Runx2 motifs associated withgenes significantly (fold change ≥1.5, FDR <0.05) downregulated (Down), upregulated (Up), or unchanged (Non-responsive) upon shRunx2-treatment.Conservation of Runx2 motifs was compared between shRunx2 downregulated and upregulated genes and length-matched non-responsive genes(Non-responsive, Down and Up, respectively) and statistical significance determined (**P< 0.05) by Kolmogorov-Smirnov test.


Table 1 Top 15 genes upregulated or downregulated duringshRunx2 treatment when compared to scramble shRNA

Down Up

Symbol Fold change Symbol Fold change

Phex 0.26 I830012O16Rik 4.73

H19 0.36 Ifi27l2a 3.83

Ibsp 0.46 Oas2 3.78

Mmp13 0.46 D14Ertd668e 3.62

Mid2 0.49 Ifit3 3.37

Edil3 0.50 Gm12250 3.27

Tnfrsf19 0.53 Ifit1 3.25

Bglap2 0.56 Ly6c1 3.24

C87414 0.56 Usp18 3.09

Gpr116 0.56 BC023105 3.05

Pdlim1 0.57 Iigp1 3.05

Hs3st3a1 0.57 Ddx60 2.88

Id3 0.58 Gm4951 2.87

Chac1 0.58 Ly6a 2.87

Il33 0.58 Ly6c2 2.85


Mmp13, which are known to be activated by Runx2 [5].Notably, the genes most upregulated by shRunx2 have notbeen characterized as Runx2 targets (Table 1) except forUsp18, which encodes a protein involved in the ubiquitindegradation pathway [28].We characterized the relationship between Runx2 oc-

cupancy and genes affected by shRunx2 knockdown,compared to non-responsive genes. The number ofpeaks, genomic distribution of peaks, and fold change ofpeak signals were compared among the gene groups atdays 0 and 9 (Figure 5B-D). Genes that were downregu-lated by shRunx2 at both days 0 and 9 had more Runx2peaks when compared to the control non-responsivegenes (Figure 5B). It should be noted, however, thatshRunx2 downregulated genes are longer than the up-regulated genes (Additional file 10). This finding wasalso observed when we included the Runx2 binding pro-files at day 28 for Runx2-regulated genes (Additional file10). In contrast, the shRunx2 upregulated genes that ap-peared on day 9 had fewer Runx2 peaks compared tocontrol genes (Figure 5B). There were also significantdifferences in overall peak distribution between shRunx2responsive genes and control (Figure 5C). Genes that weredownregulated tended to exhibit more intronic and inter-genic enrichment of Runx2 peaks, while shRunx2 upregu-lated genes were strongly enriched in intergenic butreduced in intronic binding. We further examined the dis-tribution of Runx2 peaks in up- and down-regulated genesas a function of changes in Runx2 binding during differen-tiation from days 0 to 9 (Figure 5D). The fold change inday 9/day 0 peak signals showed increased Runx2 binding

predominantly at upstream and promoter regions for theshRunx2 downregulated genes; in shRunx2 upregulatedgenes, Runx2 binding diminished in the introns andshowed no significant change in promoter regions. There-fore, shRunx2 downregulated and upregulated genes ex-hibited distinct preferences for Runx2 binding in genomicloci as reflected by peak distributions in Figure 5B,C, andin relation to differentiation (Figure 5D).To complement the above analysis examining the

genome-wide distribution of Runx2 responsive peaks, weused the PeaksToGenes program [29] to determine theenrichment of day 9 Runx2 signals at defined intervalswithin and surrounding the gene bodies (Figure 5E). ForshRunx2 downregulated genes, Runx2 binding had thestrongest enrichment surrounding TSSs, which includesproximal promoter, 5′ UTR, exons and introns all withinfive contiguous deciles (Figure 5E; Additional file 11).This analysis demonstrated that, of genes affected byshRunx2, 66.1% of upregulated genes have a low levelof Runx2 binding (cluster V in Figure S5B in Additionalfile 11) but most downregulated genes (69.1%) havehigher level intensities of Runx2 binding (Figure S5A,Bin Additional file 11). This finding shows that Runx2 re-sponsive genes at day 9 (shRunx2/Scr) are primarily reg-ulated by Runx2 surrounding TSSs.As another computational analysis, EMBER (Expect-

ation Maximization of Binding and Expression pRofiles)[30] was used to relate measured changes in gene ex-pression to the spectrum of Runx2 occupancy observedduring osteoblast differentiation (Figure 3B,C; Additionalfile 12). In analogy with discovering a sequence motiffrom a collection of functionally related DNA sequences,EMBER optimizes an ‘expression pattern’ from a collec-tion of genes related by patterns of transcription factorbinding and uses this motif to determine which genesare regulatory targets of the transcription factor (detailsof EMBER are summarized in Additional file 3). Usingthis approach, we discovered and compared expressionpatterns from different sets of Runx2 binding regions(Figure 3). The Runx2 peaks were partitioned into 42subsets of Runx2 binding regions (7 clusters with 6 gen-omic location categories) and it was observed that allgroups of Runx2 binding show a correlative relationshipto gene expression during osteoblast differentiation(Additional file 12). We noted, however, that intronicRunx2 binding regions - for all time-dependent patternsof Runx2 occupancy - are less informative than othergroups, indicating that intronic Runx2 binding may beless useful than binding at other class elements as a pre-dictor of gene regulation.Finally, downregulated and upregulated genes were

compared on the basis of their evolutionary conservation(Figure 5F). Functional genomic elements are often char-acterized by conservation, which has been used to guide


the prediction of transcription factor binding sites[31,32]. This finding has been shown to distinguish func-tionally verified from un-verified binding sites in a large-scale study of transcription factor binding site functionon human promoters [33]. For each ChIP-Seq peak, aRunx motif was used to identify the single most likelyRunx2 binding site and the mean phyloP score [34] (forconservation among 30 vertebrate species) across thebinding sites was computed. For each gene, phyloPscores were averaged among all ChIP-Seq peaks thatwere associated with individual genes to give an averagemeasure of Runx2 conservation. We found that genesdownregulated by shRunx2 had Runx2 binding se-quences that were more conserved than those in upreg-ulated or non-responsive genes (Figure 5F), suggestingthat Runx2 regulation may be more evolutionarily con-served in genes that are activated by Runx2 duringosteoblastogenesis.Taken together, the complementary methods described

above suggest that Runx2 employs different mechanisms toregulate gene expression: 1) shRunx2 downregulated genesshow increased Runx2 binding at promoter and far up-stream regions; and 2) based on EMBER, intronic bindingdoes not imply Runx2-mediated gene regulation to the samedegree as Runx2 binding at promoter, exon and upstreamregions. Thus, during the normal course of osteoblast differ-entiation, Runx2-activated genes (shRunx2 downregulated)are regulated through both promoter and non-promoter re-gions; and regulation of Runx2 repressed genes (shRunx2upregulated) also occurs in promoter and other genomic re-gions, but with less Runx2 binding.

Novel genes regulated by Runx2 through distinctregulatory elementsAmong potential Runx2 targets, we identified Tnfrsf19(tumor necrosis factor receptor superfamily, member19), which is involved in bone formation as a Wnt-responsive regulator of mesenchymal stem cell com-mitment to osteoblastic lineage [35]. Tnfrsf19 exhibitedenrichment of Runx2 binding to the promoter as well asintronic regions (Additional files 6 and 13). During dif-ferentiation of osteoblasts, Tnfrsf19 mRNA levels in-creased dramatically more than 100-fold from day 0 today 9 (Figure S7A in Additional file 13). Consistent withAffymetrix data, depletion of Runx2 resulted in signifi-cant decreases in Tnfrsf19 expression, indicating directRunx2 regulation (Figure S7B in Additional file 13). Wefound that Runx2 occupancy was increased at intronicand promoter regions of the Tnsrsf19 locus during osteo-blast differentiation (Figure S7C in Additional file 13).Adamts4 and Crabp2 were selected for further ana-

lyses based on their responsiveness to shRunx2 (withfold change cutoff of 1.3; Additional file 14) and theirRunx2 binding patterns predominantly in non-promoter

regions. To test the functionality of Runx2 binding at theseputative regulatory regions, we cloned non-promoter Runx2binding regions and measured transcriptional activityby luciferase reporter assay. Adamts4 is expressed inosteoblasts and osteocytes and encodes an enzyme thatdegrades aggrecan [36,37]. Runx2 exhibits multiple peaksacross the Adamts4 locus, with increased occupancyduring osteoblast differentiation (Figure 6A; Additionalfile 6). Adamts4 expression during osteoblast differenti-ation was increased at day 9 and remained steady to day28 (Figure 6B). Knockdown of Runx2 suppressed the ex-pression of Adamts4 (Figure 6C) by 40%, supportingAdamts4 as a direct target of Runx2 during osteoblasto-genesis. We characterized the functional activity of twoprominent Runx2 binding regions, peak A in intron 1 andpeak B in the last exon (Figure 6A). The peak A regionincreased luciferase activity in MC3T3-E1 cells by over20-fold. In contrast, peak B functioned as a suppressorof luciferase activity (Figure 6D). These results indicatethat the two Runx2 sites can function as a positive andnegative regulator of Adamts4; however, the weaker,negative regulation by peak B on the luciferase reportermay be due to the lack of a native chromatin context.The large increase in luciferase activity observed frompeak A is consistent with a previous study that demon-strated Runx2 upregulates ADAMTS4 in human SW1353chondrosarcoma cells [38]. Here we established in osteo-blasts that Runx2-mediated upregulation of Adamts4 canoccur at non-promoter regulatory elements, and is not re-stricted to the proximal promoter region as previouslyshown [38].Crabp2 is a cytoplasmic retinoic acid binding protein

previously reported to be upregulated during osteoblasto-genesis [39]. We observed that Runx2 constitutively occu-pies the Crabp2 locus in the first intron, while bindingincreases upstream and downstream of the Crabp2 genebody during differentiation (Figure 6E; Additional file 6).Crabp2 was upregulated during MC3T3-E1 cell differenti-ation and knockdown of Runx2 decreased Crabp2 mRNAlevel (Figure 6F). In proliferating cells, peak regions C, Dand E demonstrated minimal luciferase reporter activ-ity. However, in differentiating cells, all peaks exhibited asignificant increase of luciferase activity, with peak region Dshowing a more than 30-fold activation (Figure 6G). Theknockdown of Runx2 reduced luciferase activity (Figure 6H),further supporting the function of these regions in mediat-ing Runx2 regulation of Crabp2.These findings of novel genes bound and regulated by

Runx2 through different types of genomic elements sup-port an emerging concept that non-promoter elementscan regulate gene transcription. Our results also indicatethat Runx2 mediates complex fine-tuning of gene expres-sion in osteoblasts by both activating and repressing regu-latory elements that are located in non-promoter regions.

peak E

3 kb

Chr3: 87,747,500 87,765,000Crabp2

Runx2 Consensus

10

0

10

0

10

0

_

_

_

_

_

_

0

2

4

6

8

10

12

14


Crabp2

0

0.4

0.8

1.2

Crabp2

Rel

ativ

e m

RN

A e

xpre

ssio

n

peak C D E0

5

10

15

20

25

30

35

Rel

ativ

e lu

cife

rase

act

ivity

Day 7Day 0

Rel

ativ

e m

RN

A e

xpre

ssio

n

0

1

2

3

4

5

6

peak C E D 0

0.2

0.4

0.6

0.8

1

WT

Rel

ativ

e lu

cife

rase

act

ivity

peak Cpeak D

F G H

Runx2 Consensus

5 kb

Chr1: 173,175,000 173,195,000Ndufs2 Adamts4

35

0

35

0

35

0

_

_

_

_

_

_

Adamts4


1

2

3

4

5

6

Rel

ativ

e m

RN

A e

xpre

ssio

n

peak A peak B

0

0.2

0.4

0.6

0.8

1

1.2

Rel

ativ

e m

RN

A e

xpre

ssio

n

Adamts4

Scr

B C D

0

5

10

15

20

25

0

0.2

0.4

0.6

0.8

1

peak A B

Rel

ativ

e lu

cife

rase

act

ivity

1.230

*

*

*

******

*

*

**

Day 7Day 0

shRunx2

ScrshRunx2

shRunx2

day 0

day 9

day 28

day 0

day 9

day 28

input_

_

35

0

A

E

input_

_

10

0

Figure 6 (See legend on next page.)


(See figure on previous page.)Figure 6 Non-promoter association of Runx2 regulates novel targets Adamts4 and Crabp2. (A) Increasing Runx2 enrichment was observedin the first intron (peak A, boxed region) and last exon (peak B, boxed region) of the Adamts4 locus during osteogenic differentiation. (B) Adamts4mRNA levels (normalized to Hprt1) were significantly upregulated (P< 0.05) in differentiating MC3T3-E1 cells (days 9 to 28) when compared to proliferatingcells (day 0). (C) Runx2 knockdown (shRunx2) decreases Adamts4 expression, compared to a scrambled shRNA (Scr) control (**P< 0.05). (D) DNA sequencesidentical to peak regions A and B were cloned into individual pGL2-SV40-Luc reporters and relative luciferase activity was measured in transfected MC3T3-E1 cellsand significant increases and decreases (*P<0.01) in luciferase activity were observed for peak A and B reporters, respectively. (E) Increasing Runx2 enrichmentwas observed up- (peak C) and downstream (peaks D and E) of the Crabp2 locus during osteogenic differentiation. (F) Crabp2 expression increases duringdifferentiation. Knockdown of Runx2 (by shRunx2) significantly reduces (**P<0.05) the endogenous expression of Crabp2 (right panel). (G) DNA sequencesidentical to peak regions C, D and E were cloned into individual pGL2-SV40-Luc reporters and relative luciferase activity was measured in transfected MC3T3-E1cells at days 0 and 7 (*P<0.01). (H) Runx2 knockdown by inducible shRNA results in a significant decrease (**P<0.05) of luciferase activity mediated by peak Cand E regions and a downward trend (P=0.08) in luciferase activity regulated by peak D region. Statistical significance for all experiments was determined byStudent’s t-test from mean±SEM from three biological replicates.


DiscussionThrough comprehensive genomic analysis of Runx2 byChIP-Seq, we describe widespread Runx2 binding through-out the genome of differentiating osteoblasts. In additionto Runx2 interaction at promoters, we find Runx2 bindingin non-promoter regions regulating novel targets that aresilenced or expressed at different stages of osteoblast dif-ferentiation. Runx2 peaks cluster into temporal and func-tional categories associated with genes in a broad range ofcellular programs, including bone development, negativeregulation of proliferation, active matrix formation andmechanisms for mineral deposition that reflect the progres-sion of osteogenesis. Our data have identified new Runx2-regulated genes (Tnfrsf19, Adamts4, Crabp2, and Ezh2) thathave established roles in bone formation, and more import-antly, have extended the understanding of Runx2-mediatedgene regulation to a broader range of cellular functions dur-ing osteoblast differentiation.

Runx2 binding patterns identify stage-dependent osteogenic programsTime-dependent Runx2 binding patterns underlie the dy-namic gene regulation by Runx2 during osteoblastogene-sis. We identified a large number of peaks, consistent withthe increasing protein levels of Runx2 from the earlyosteoprogenitor to the mature osteoblast/osteocyte. Theresults from binary clustering, together with subsequentGO term analyses by GREAT, identified many categoriesassociated with skeletal development and bone homeosta-sis. These findings support our initial hypothesis that dis-tinct binding patterns of Runx2 at different stages ofosteoblastogenesis have novel functional implications.Our clustering analysis partitioned Runx2 peaks into two

main categories: less variable steady-state binding (cluster1, ubiquitous peaks) and more dynamic binding groups(clusters 2 to 7, stage-specific clusters). Steady-state bindingof Runx2 persists from proliferative pre-osteoblasts to dif-ferentiating osteoblasts with binding signals plateauing dur-ing matrix deposition and mineralization stages. Peaks inthis category represent genes related to housekeepingprocesses such as protein folding, negative regulation of

mitotic cell cycle, and mRNA catabolism and processes notpreviously related to Runx2 (Additional file 4). The GOterms associated with stage-specific clusters include nega-tive regulators of other cell lineages (that is, fat and smoothmuscle cells) as well as positive regulators of osteogenesis.Thus, dynamic Runx2 binding primes, enhances and stabi-lizes the osteoblast phenotype as well as suppresses non-osteoblast lineages. Many Runx2 bound genes in cluster 4(days 9 & 28, differentiation) have been demonstrated tocontribute to in vivo bone formation [40]. Thus, the gen-omic profiling of Runx2 binding in our in vitro model sys-tem is consistent with the known properties of Runx2in bone formation. More importantly, our profiling re-veals pathways previously unknown to be controlled byRunx2 underlying biological mechanisms of general cellularprocesses.

Runx2 binding functions at both non-promoter andpromoter regionsOnly a small proportion of sequence-specific transcriptionfactors, such as Myc, have narrow distributions of bindingaround proximal promoters of genes [41-44]. In contrast,non-promoter binding is recognized for other transcrip-tion factors, including STAT1, RUNX1, ERα, CTCF, andHNF4α [41,45-49]. In a previous study, overexpression ofRunx2 in prostate cancer cells revealed extensive non-promoter binding [50]. In our study, endogenous Runx2binding across the genome was characterized during os-teoblastogenesis. We found that over 70% of Runx2 occu-pancy was localized to non-promoter regions (intergenic,intron, exon, TTS, and upstream regions that consti-tute the bulk of the genome) during the differentiation ofosteoblasts. From days 0 to 9 there was a two-fold in-crease in the number of non-promoter peaks, indicating afunctional association with the differentiation process.Runx2-dependent regulation through non-promoter peaksaround the Adamts4 and Crabp2 genes provided directevidence that non-promoter binding of Runx2 controlsgene expression.Although we have demonstrated the importance of

non-promoter Runx2 binding events, Runx2 peaks at


promoter regions have critical regulatory roles as well.In our data, Runx2 occupancy has the highest enrich-ment at promoter regions when compared with othergenomic locations. Examples of this regulatory modecan be seen in some well-characterized Runx2 targetssuch as Bsp and Ocn, in line with established evidencethat Runx2 can regulate these genes in a promoter-dependent manner [51-54]. The genes downregulatedupon Runx2 silencing also displayed a clear enrichmentof Runx2 signal at promoter regions (Figure 5B), exem-plified in Tnfrsf19.Gene expression in different biological settings is influ-

enced by higher order three-dimensional chromatincomplexes that involve looping of promoter and non-promoter elements, blurring the distinction of definedregulatory regions [55,56]. It is plausible that someRunx2 peaks in promoter and non-promoter regionsmay serve as nucleation sites for modifications of chro-matin structures necessary for gene expression. Further-more, transcription factors can interact with RNApolymerase II, CTCF, and other factors via higher-orderchromatin conformations [55,57,58]. During osteoblasto-genesis, Runx2 exhibited a distribution pattern amonggenomic elements similar to that of CTCF, suggestingthat, like CTCF, Runx2 may have a functional role thatextends beyond direct regulation of transcription. Con-sistent with this idea, Runx2 is able to form discerniblefoci associated with the nuclear matrix: a nuclear frame-work for organizing higher order chromatin structures.Runx2 truncated of the NMTS (nuclear matrix-targetingsignal) domain results in diminished nuclear matrix as-sociation and disrupted expression of Runx2 target genes[59]. It is also noteworthy that the genes upregulatedupon Runx2 knockdown have preferential Runx2 bind-ing in intergenic regions, indicating that distal elementsmay have a regulatory role through long-range interac-tions. In addition, Runx2 binding at distal regions maycontribute to chromatin remodeling through interactingwith chromatin-modifying enzymes, as has been welldocumented at regulatory elements in osteoblasts andother cell models [12,13].

The complexities of Runx2 binding and transcriptionalregulationRunx2 displays complex binding patterns similar to otherlineage-specific transcription factors, such as PPAR-γ,MyoD, and GATA3 [16-19]. By systematic annotation ofRunx2 peaks, multiple integrative analyses of gene expres-sion combined with Runx2 binding profiles and directexperimental validation of individual targets, we definedRunx2 binding with biological outcomes during osteoblastdifferentiation. These analyses revealed that, for a smallset of genes, the enrichment and binding patterns ofRunx2 were indicators of gene expression. Genes that have

decreased expression in the absence of Runx2 (byshRunx2 treatment) have a greater number of Runx2peaks and greater fold change of peak signal at promoterand non-promoter regions when compared to non-responsive and gene length-matched controls. In contrast,genes upregulated by Runx2 knockdown tend to havefewer Runx2 peaks and smaller relative fold changes inpeak signals when compared to controls. One notablefinding that arose from our analysis was that genes thatwere downregulated in the absence of Runx2 had both agreater evolutionary conservation of Runx2 binding sitesand tended to be longer than non-responsive and upregu-lated genes. It is unclear why this particular subset ofgenes (that is, genes normally activated by Runx2) wouldretain these features throughout evolution; however, this isan interesting point for future investigations. Although wedemonstrated that Runx2 binding influences gene expres-sion, a proportion of Runx2 peaks were found to have nodirect function in transcriptional control of genes. This maybe due to efficient but not complete knockdown of Runx2by viral-mediated shRNA. Alternatively, similar to othertranscription factors, many binding regions were found tobe nonfunctional in transactivating luciferase reporters [33].This finding suggests that binding of transcription factorsmay have a distinct function other than direct control ofgene expression, consistent with previously described non-transcriptional functions and genome-organizing capabilitiesof Runx2 [59,60].

ConclusionsOur findings provide a new level of understanding of theRunx2-mediated transcription program as defined bygenome-wide Runx2 binding essential for osteoblasto-genesis. Our data support that Runx2 functions at pro-moter and non-promoter regions at both previouslyknown and novel targets. The impact of our study exam-ining the global occupancy of endogenous Runx2 in dif-ferentiating osteoblasts sets a framework for novelmechanisms underlying bone biology.

Materials and methodsCell cultureThe calvaria-derived preosteoblast cell line MC3T3-E1(Subclone 4) was obtained from ATCC (Manassas, VA,USA) and maintained in ascorbic acid-free alpha-MEM(Hyclone, Novato, CA, USA) supplemented with 10% fetalbovine serum (FBS; Hyclone), 2 mM L-glutamine, 100 U/mlof penicillin and 100 μg/ml streptomycin (Pen/Strep, Invi-trogen, Carlsbad, CA,). To induce osteogenic differentiation,complete alpha-MEM was supplemented with 280 μMascorbic acid and 10 mM beta-glycerophosphate (SigmaAldrich, St. Louis, MO, USA). Cells were maintained at 37°Cin a humidified 5% CO2 environment and media replacedevery 2 to 3 days for the duration of all experiments.


RNA isolation and quantitative PCRTotal RNA was isolated using Trizol reagent (Invitrogen)according to the manufacturers’ specifications. Total cel-lular RNA treated by DNaseI (Zymo, Irvine, CA, USA)was primed with random hexamers and reverse tran-scribed into cDNA using Superscript First-strand cDNASynthesis Kit (Invitrogen) according to the manufac-turer’s instructions. Gene expression was determined byquantitative real time PCR (qPCR) using iQ™ SYBR GreenPCR Master Mix (BioRad, Hercules, CA, USA) in an ABIPrism 7300 thermocycler (Applied Biosystems, FosterCity, CA, USA). For each gene, the expression level wasnormalized to that of Hprt1 using 2-ΔΔCT method. Experi-ments were performed in triplicate and results are pre-sented as mean values ± SEM. Primers for qPCR reactionswere designed by FoxPrimer [61,62], and are available inAdditional file 15.

Runx2 knockdown and gene expression profilingLentiviruses carrying Runx2-shRNA and previously de-scribed control Scramble-shRNA [63] were used to in-fect MC3T3-E1 cells. Infected cells were subsequentlydetected by green fluorescent protein, sorted and grownto 90% confluency followed by osteogenic differentiationfor 9 days. Knockdown experiments were performed inthree biological replicates.Microarray samples were handled following the manu-

facturers’ recommended protocols (Affymetrix, SantaClara, CA, USA). Briefly, RNA isolated from MC3T3-E1cells (day 0 and 9 scramble shRNA, and day 9 Runx2shRNA) were reversely transcribed into cDNA using WTExpression Kit (Ambion, Austin, TX, USA), labeled andfragmented with GeneChip WT Terminal Labeling andControls Kit. Labeled cDNAs were then hybridized toGeneChip Mouse Gene 1.0 ST Array rev.4 using a Gene-Chip Hybridization, Wash, and Stain Kit. Hybridizationsignals were obtained by GeneChip Scanner (Affymetrix).Microarray data were analyzed using Bioconductor (ver-sion 2.11) packages affy and limma in R (version 2.15.1)[64-67]. Briefly, after performing RMA (robust multichipaverage) normalization of microarray expression levelsand filtering, differential expression was detected using aBayesian moderated t-test. The Benjamini-Hochberg FDR[68] was applied to correct for multiple testing. The Affy-metrix expression profiles were annotated to RefSeq genes[23]. These expression data have been deposited in theGene Expresion Omnibus (GEO) database under acces-sion number GSE53982.

Western blotNuclei extracts were prepared from MC3T3-E1 cellsusing a protocol modified from Dignam et al. [69]. Pri-mary antibodies and dilutions were: mouse anti-Runx2monoclonal (Clone 8G5, MBL International, Woburn,

MA, USA; 1:1,000); rabbit anti-H3 3H1 monoclonal(Cell Signaling, Danvers, MA, USA; 1:2,000). HRP-conjugated secondary antibodies and dilutions were: goatanti-mouse IgG (Santa Cruz, Dallas, TX, USA; 1:3,000);goat anti-rabbit IgG (Santa Cruz; 1:3,000). Detection ofHRP was performed using a Western Lightning Plus Kit(Perkin Elmer, Waltham, MA, USA) followed by expos-ure onto Biomax Light film (Kodak, Rochester, NY,USA).

Chromatin immunoprecipitation and high-throughputsequencingAt days 0, 9, and 28 of differentiation, approximately 1 ×108 MC3T3-E1 cells were washed with PBS (phosphate-buffered saline) and then fixed on a plate with 1% for-maldehyde for 8 minutes to crosslink DNA-protein com-plexes. The fixed cells were washed with ice-cold PBS,harvested, and pelleted. Nuclei extraction was performedusing a protocol modified from Dignam et al. [69]. Iso-lated nuclei were sonicated using a Misonix S-4000ultrasonic sonicator to obtain sheared chromatin rangingfrom 0.2 kb to 0.6 kb. Sheared chromatin was usedfor immunoprecipitation with Runx2 antibody (M-70,Santa Cruz) [70] or immunoglobulin G (IgG) (12-370,Millipore, Billerica, MA, USA) followed by purificationusing Protein-G Dynabeads (Invitrogen). Precipitated chro-matin was washed with solutions of increasing salt con-centration, and eluted and subsequently uncrosslinked at65°C. DNA was recovered by phenol/chloroform/isoa-myl alcohol extraction followed by ethanol precipita-tion. Libraries of purified DNA were generated usingIllumina SR adapters (Illumina, San Diego, CA, USA)following the manufacturer’s manual. DNA librarieswere selected for inserted fragments of 200 ± 50 bp,and single-end 36 base reads were generated on an Illu-mina Genome Analyzer II at the UMASS Deep-Seqcore facility. Base calls and sequence reads were gener-ated by Illumina CASAVA software (version 1.6; Illu-mina). Two independent biological repeats of Runx2ChIP-Seq libraries were prepared for each time point,and two Input libraries were prepared with sonicatedDNA from day 9 MC3T3-E1 cells.

Analysis of ChIP-Seq dataSingle-end 36 base sequences from Runx2 ChIP-Seq andinput libraries were mapped to the mouse genome (as-sembly mm9) using Bowtie (version 0.12.8) [71]. Runx2peaks and read counts were determined by MACS (ver-sion 1.4.1) [22] using default settings. Runx2 peaks thatwere significant at the P < 10-10 level were retained forsubsequent analysis. For each of the three time points inour data, read counts were normalized to 10 millionreads. The UCSC genome browser [72] was used tovisualize Runx2 peaks.


Runx2 binding regions were classified on the basis ofgenomic location categories and annotated to knownRefSeq genes [23]. Runx2 peaks were grouped into sevenclusters based on the presence or absence of a peak.Runx2 peaks in each cluster were analyzed for GO termsusing GREAT (version 2.0.2), using default associationrules between ChIP-Seq peaks and annotated genes [73].The conservation of Runx2 motifs associated with thegenes responsive to shRunx2 were determined usingphyloP [34]. Statistical significance was determined usingKolmogorov-Smirnov test.PeaksToGenes [29,62] (Additional file 3) was used to

test Runx2 binding in relation to the genes from micro-array analysis. PeaksToGenes defines genomic intervalsrelative to all RefSeq genes, and in each window uses anon-parametric Wilcoxon rank sum test to calculate theprobability and binding frequency. The comparisonswere individually made between each responsive groupand the non-responsive group to Runx2 shRNA. Runx2binding and Runx2-mediated transcription control werealso evaluated by EMBER [30]. Analogous to discoveringa sequence motif from functionally related DNA se-quences [24], EMBER optimizes an expression patternfrom a collection of genes’ expression data related byprofiles of transcription factor binding and uses this in-formation to determine which genes are potential regu-latory targets of the transcription factor. For a detaileddescription of the PeaksToGenes and EMBER analysis,please refer to Additional file 3.Functional genomic elements can be characterized by

evolutionary conservation and have been shown to dis-tinguish functionally verified from unverified transcrip-tion factor binding sites on human promoters [33]. Foreach ChIP-Seq peak (combined from day 0, 9, and 28datasets), a Runx motif was used to identify the singlemost likely Runx2 binding site and the mean phyloPscore [34] (for conservation among 30 vertebrate spe-cies) across the binding sites was computed. In orderto compare conservation among genes that fell intodifferent regulatory groups (shRunx2 downregulated, up-regulated and non-responsive genes) based on our ex-pression microarray measurements, phyloP scores wereaveraged among all ChIP-Seq peaks that were associatedwith individual genes to give an average measure ofRunx2 conservation for each gene. To minimize theeffect of gene length on Runx2 binding analysis, wecompared the conservation of shRunx2-downregulated and -upregulated genes to randomly sampled, length-matchednon-responsive genes. Statistical significance was deter-mined by pair-wise conservation comparisons usingKolmogorov-Smirnov test.The raw sequences and peak-related files in BED and WIG

formats representing processed data have been deposited inthe GEO database under accession number GSE54013.

Luciferase assays and plasmid reportersSelected Runx2 peak regions were cloned with MluIcombined with BglII or XhoI into a pGL2-SV40-Luc re-porter (Promega, Fitchburg, WI, USA). Primers used incloning are listed in Additional file 16. A pGL2-SV40-Luc reporter with minimum SV40 promoter was used asmock control. Reporter plasmids with Runx2 peak regionsand pGL2-SV40-Luc empty vector were co-transfectedwith pcDNA3.1-Runx2-WT or pcDNA3.1-EV into MC3T3cells. Transfected cells were then differentiated for 7 daysbefore luciferase activities were determined by Dual-GloLuciferase Assay Kit (Promega). A second set of luciferaseassays was done in differentiating MC3T3-E1 cells stablyinfected with a doxycycline-inducible pLKO-puro-Tet-on-Rx2shRNA lentiviral vector. This vector was constructedby re-cloning a previously validated Runx2 shRNA se-quence [63] to Tet-pLKO-puro plasmid (catalog number21915, Addgene, Cambridge, MA, USA). Luciferase activ-ities from MC3T3-E1 cells treated with or without 2.5μg/ml doxycycline were examined at day 7 after differenti-ation (Figure 6).

Additional files

Additional file 1: Table S3. Summary of MC3T3 Runx2 ChIP-Seq. This tablelists the read numbers and genome coverage of Runx2 ChIP-Seq libraries.

Additional file 2: Table S4. Distribution patterns of Runx2 peaks acrossgenomic locations. This table is related to Figure 2.

Additional file 3: Detailed description of ChIP-PCR, ChIP-Seq withbioinformatics analysis, and supplemental figure legends.

Additional file 4: Table S5. GREAT Gene Ontology terms. This tablecontains the top GO terms assigned by GREAT to clusters 1 to 7 definedin Figure 3A.

Additional file 5: Figure S1. GO term analysis from GREAT for clusters1 to 3 and 5 to 7 in Figure 3A. This is related to Figure 3.

Additional file 6: Figure S8. Validation of Runx2 peaks by ChIP-PCR.This figure is related to Figures 4, 5 and 6.

Additional file 7: Figure S2. Runx2 peaks associated with Bsp geneduring differentiation. This figure is related to Figure 4.

Additional file 8: Table S6. Detailed annotation of Runx2 peaks. Thistable is a meta-spreadsheet of Runx2 peaks annotated to RefSeq genes.

Additional file 9: Figure S3. Validation of Runx2 knockdown in MC3T3cells. This figure is related to Figure 5.

Additional file 10: Figure S4. Additional characteristics of Runx2binding in shRunx2 responsive genes. This figure is related to Figure 5.

Additional file 11: Figure S5. PeaksToGenes analysis of Runx2 occupancyin Runx2 shRNA-responsive genes. This figure is related to Figure 5E.

Additional file 12: Figure S6. EMBER analyses of Runx2 binding in thegenes differentially regulated by Runx2 knockdown. This figure is relatedto Figure 5.

Additional file 13: Figure S7. Validation of novel Runx2 target Tnfrsf19.This figure is related to Figure 6.

Additional file 14: Table S7. Genes responsive to Runx2 shRNA. Thistable lists the genes that are up- or downregulated by shRunx2 treat-ment in MC3T3 cells differentiated for 9 days.

Additional file 15: Table S1. qPCR and ChIP-PCR primers.

Additional file 16: Table S2. Cloning primers. This table contains theprimers used for plasmid construction.

http://www.biomedcentral.com/content/supplementary/gb-2014-15-3-r52-S1.pdf



http://www.biomedcentral.com/content/supplementary/gb-2014-15-3-r52-S4.xlsx














Abbreviationsbp: base pair; ChIP: chromatin immunoprecipitation; FDR: false discovery rate;GEO: Gene Expresion Omnibus; GO: Gene Ontology; qPCR: quantitative realtime polymerase chain reaction; shRNA: short hairpin RNA; TSS: transcriptionstart site; TTS: transcription termination site; UTR: untranslated region, SEM,standard error of mean.

Competing interestsThe authors declare that they have no competing interests.

Authors’ contributionsHW and JARG conceived and designed the experiments. HW performed theexperiments. TWW, JRD, and HW performed bioinformatics data analysis. HW,JARG, TWW, and JRD interpreted the data with insights from PWLT, JBL, JLS,AVW, and GSS. HW, TWW, JARG, JRD, PWLT, JBL, and JLS wrote the paper withhelp from AVW and GSS. All authors read and approved the final manuscript.

AcknowledgementsThis work is supported by grants R37 DE012528 and R37 DE012528-24S1 toJane B Lian, and grant R01 AR039588 to Gary S Stein. We thank Dana Frederickfor her excellent technical support, and Jill Moore, an undergraduate at Universityof Massachusetts at Amherst, for her bioinformatics assistance. We are grateful toEllen LW Kittler and the Deep-sequencing Core Facility at UMass Medical Schoolfor their excellent services and technical support. We thank Phyllis Spatrick andthe Genomics Core Facility at UMass Medical School for their excellent servicesand supports. We thank Jeffrey P Bond in the Department of Microbiology andMolecular Genetics of University of Vermont for his suggestions on the manuscript.

Author details1Department of Biochemistry, University of Vermont College of Medicine andVermont Cancer Center, 89 Beaumont Avenue, Burlington, VT 05405, USA.2Department of Cell & Developmental Biology, University of MassachusettsMedical School, 364 Plantation Street, Worcester, MA 01655, USA. 3Currentaddress: Center for Computational Molecular Biology, Department ofMolecular Biology, Cell Biology & Biochemistry, Brown University, 115Waterman Street, Providence, RI 02912, USA. 4Current address: Departmentof Computer Science, Brown University, 115 Waterman Street, Providence, RI02912, USA. 5Current address: Departments of Orthopedic Surgery andBiochemistry & Molecular Biology, Mayo Clinic, Medical Sciences Building3-69, 200 First Street SW, Rochester, MN 55905, USA.

Received: 27 August 2013 Accepted: 21 March 2014Published: 21 March 2014

References1. Komori T: Signaling networks in RUNX2-dependent bone development.

J Cell Biochem 2011, 112:750–755.2. Lian JB, Stein GS, Javed A, van Wijnen AJ, Stein JL, Montecino M, Hassan

MQ, Gaur T, Lengner CJ, Young DW: Networks and hubs for thetranscriptional control of osteoblastogenesis. Rev Endocr Metab Disord2006, 7:1–16.

3. Long F: Building strong bones: molecular regulation of the osteoblastlineage. Nat Rev Mol Cell Biol 2012, 13:27–38.

4. Otto F, Kanegane H, Mundlos S: Mutations in the RUNX2 gene in patientswith cleidocranial dysplasia. Hum Mutat 2002, 19:209–216.

5. Hecht J, Seitz V, Urban M, Wagner F, Robinson PN, Stiege A, Dieterich C,Kornak U, Wilkening U, Brieske N, Zwingman C, Kidess A, Stricker S, MundlosS: Detection of novel skeletogenesis target genes by comprehensiveanalysis of a Runx2(-/-) mouse model. Gene Expr Patterns 2007, 7:102–112.

6. Vaes BL, Ducy P, Sijbers AM, Hendriks JM, van Someren EP, de Jong NG, vanden Heuvel ER, Olijve W, van Zoelen EJ, Dechering KJ: Microarray analysison Runx2-deficient mouse embryos reveals novel Runx2 functions andtarget genes during intramembranous and endochondral bone formation.Bone 2006, 39:724–738.

7. Jeon MJ, Kim JA, Kwon SH, Kim SW, Park KS, Park SW, Kim SY, Shin CS:Activation of peroxisome proliferator-activated receptor-gamma inhibitsthe Runx2-mediated transcription of osteocalcin in osteoblasts.J Biol Chem 2003, 278:23270–23277.

8. Zhang YY, Li X, Qian SW, Guo L, Huang HY, He Q, Liu Y, Ma CG, Tang QQ:Down-regulation of type I Runx2 mediated by dexamethasone isrequired for 3T3-L1 adipogenesis. Mol Endocrinol 2012, 26:798–808.

9. Gersbach CA, Byers BA, Pavlath GK, Garcia AJ: Runx2/Cbfa1 stimulatestransdifferentiation of primary skeletal myoblasts into a mineralizingosteoblastic phenotype. Exp Cell Res 2004, 300:406–417.

10. Schroeder TM, Jensen ED, Westendorf JJ: Runx2: a master organizer ofgene transcription in developing and maturing osteoblasts. Birth DefectsRes C Embryo Today 2005, 75:213–225.

11. Zaidi SK, Young DW, Montecino MA, Lian JB, van Wijnen AJ, Stein JL, SteinGS: Mitotic bookmarking of genes: a novel dimension to epigeneticcontrol. Nat Rev Genet 2010, 11:583–589.

12. Bradley EW, McGee-Lawrence ME, Westendorf JJ: Hdac-mediated control ofendochondral and intramembranous ossification. Crit Rev Eukaryot GeneExpr 2011, 21:101–113.

13. Pelletier N, Champagne N, Stifani S, Yang XJ: MOZ and MORF histoneacetyltransferases interact with the Runt-domain transcription factorRunx2. Oncogene 2002, 21:2729–2740.

14. Villagra A, Cruzat F, Carvallo L, Paredes R, Olate J, van Wijnen AJ, Stein GS,Lian JB, Stein JL, Imbalzano AN, Montecino M: Chromatin remodeling andtranscriptional activity of the bone-specific osteocalcin gene requireCCAAT/enhancer-binding protein beta-dependent recruitment of SWI/SNF activity. J Biol Chem 2006, 281:22695–22706.

15. van Wijnen AJ, Stein GS, Gergen JP, Groner Y, Hiebert SW, Ito Y, Liu P, NeilJC, Ohki M, Speck N: Nomenclature for Runt-related (RUNX) proteins.Oncogene 2004, 23:4209–4210.

16. Mikkelsen TS, Xu Z, Zhang X, Wang L, Gimble JM, Lander ES, Rosen ED:Comparative epigenomic analysis of murine and human adipogenesis.Cell 2010, 143:156–169.

17. Cao Y, Yao Z, Sarkar D, Lawrence M, Sanchez GJ, Parker MH, MacQuarrie KL,Davison J, Morgan MT, Ruzzo WL, Gentleman RC, Tapscott SJ: Genome-wideMyoD binding in skeletal muscle cells: a potential for broad cellularreprogramming. Dev Cell 2010, 18:662–674.

18. Zhang JA, Mortazavi A, Williams BA, Wold BJ, Rothenberg EV: Dynamictransformations of genome-wide epigenetic marking and transcriptionalcontrol establish T cell identity. Cell 2012, 149:467–482.

19. Handoko L, Xu H, Li G, Ngan CY, Chew E, Schnapp M, Lee CW, Ye C, Ping JL,Mulawadi F, Wong E, Sheng J, Zhang Y, Poh T, Chan CS, Kunarso G, Shahab A,Bourque G, Cacheux-Rataboul V, Sung WK, Ruan Y, Wei CL: CTCF-mediatedfunctional chromatin interactome in pluripotent cells. Nat Genet 2011,43:630–638.

20. Wang D, Christensen K, Chawla K, Xiao G, Krebsbach PH, Franceschi RT:Isolation and characterization of MC3T3-E1 preosteoblast subclones withdistinct in vitro and in vivo differentiation/mineralization potential.J Bone Miner Res 1999, 14:893–903.

21. Choi JY, Lee BH, Song KB, Park RW, Kim IS, Sohn KY, Jo JS, Ryoo HM:Expression patterns of bone-related proteins during osteoblasticdifferentiation in MC3T3-E1 cells. J Cell Biochem 1996,61:609–618.

22. Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, NusbaumC, Myers RM, Brown M, Li W, Liu XS: Model-based analysis of ChIP-Seq(MACS). Genome Biol 2008, 9:R137.

23. Pruitt KD, Tatusova T, Maglott DR: NCBI reference sequences (RefSeq): acurated non-redundant sequence database of genomes, transcripts andproteins. Nucleic Acids Res 2007, 35:D61–D65.

24. Bailey TL, Boden M, Buske FA, Frith M, Grant CE, Clementi L, Ren J, Li WW,Noble WS: MEME SUITE: tools for motif discovery and searching.Nucleic Acids Res 2009, 37:W202–W208.

25. Lee BK, Iyer VR: Genome-wide studies of CCCTC-binding factor (CTCF)and cohesin provide insight into chromatin structure and regulation.J Biol Chem 2012, 287:30906–30913.

26. Portales-Casamar E, Thongjuea S, Kwon AT, Arenillas D, Zhao X, Valen E,Yusuf D, Lenhard B, Wasserman WW, Sandelin A: JASPAR 2010: the greatlyexpanded open-access database of transcription factor binding profiles.Nucleic Acids Res 2010, 38:D105–D110.

27. Wei Y, Chen YH, Li LY, Lang J, Yeh SP, Shi B, Yang CC, Yang JY, Lin CY,Lai CC, Hung MC: CDK1-dependent phosphorylation of EZH2 suppressesmethylation of H3K27 and promotes osteogenic differentiation ofhuman mesenchymal stem cells. Nat Cell Biol 2011, 13:87–94.

28. Malhotra S, Morcillo-Suarez C, Nurtdinov R, Rio J, Sarro E, Moreno M, CastilloJ, Navarro A, Montalban X, Comabella M: Roles of the ubiquitin peptidaseUSP18 in multiple sclerosis and the response to interferon-beta treatment.Eur J Neurol 2013, 20:1390–1397.

29. PeaksToGenes. [https://github.com/peakstogenes/PeaksToGenes]

https://github.com/peakstogenes/PeaksToGenes


30. Maienschein-Cline M, Zhou J, White KP, Sciammas R, Dinner AR:Discovering transcription factor regulatory targets using gene expressionand binding data. Bioinformatics 2012, 28:206–213.

31. Kellis M, Patterson N, Endrizzi M, Birren B, Lander ES: Sequencing andcomparison of yeast species to identify genes and regulatory elements.Nature 2003, 423:241–254.

32. Stark A, Lin MF, Kheradpour P, Pedersen JS, Parts L, Carlson JW, Crosby MA,Rasmussen MD, Roy S, Deoras AN, Ruby JG, Brennecke J, Hodges E, HinrichsAS, Caspi A, Paten B, Park SW, Han MV, Maeder ML, Polansky BJ, Robson BE,Aerts S, van Helden J, Hassan B, Gilbert DG, Eastman DA, Rice M, Weir M,Hahn MW, Park Y, et al: Discovery of functional elements in 12 Drosophilagenomes using evolutionary signatures. Nature 2007, 450:219–232.

33. Whitfield TW, Wang J, Collins PJ, Partridge EC, Aldred SF, Trinklein ND, Myers RM,Weng Z: Functional analysis of transcription factor binding sites in humanpromoters. Genome Biol 2012, 13:R50.

34. Siepel A, Pollard KS, Haussler D: New methods for detecting lineage-specific selection. In Research in Computational Molecular Biology, 10thAnnual International Conference, RECOMB 2006; April 2-5, 2006: Venice, Italy.Edited by Apostolico A, Guerra C, Istrail S, Pevzner P, Waterman M. Berlin:Springer-Verlag; 2006:190–205.

35. Qiu W, Hu Y, Andersen TE, Jafari A, Li N, Chen W, Kassem M: Tumornecrosis factor receptor superfamily member 19 (TNFRSF19) regulatesdifferentiation fate of human mesenchymal (stromal) stem cells throughcanonical Wnt signaling and C/EBP. J Biol Chem 2010, 285:14438–14449.

36. Sone S, Nakamura M, Maruya Y, Takahashi I, Mizoguchi I, Mayanagi H,Sasano Y: Expression of versican and ADAMTS during rat tooth eruption.J Mol Histol 2005, 36:281–288.

37. Wang K, Vishwanath P, Eichler GS, Al-Sebaei MO, Edgar CM, Einhorn TA,Smith TF, Gerstenfeld LC: Analysis of fracture healing by large-scaletranscriptional profile identified temporal relationships betweenmetalloproteinase and ADAMTS mRNA expression. Matrix Biol 2006,25:271–281.

38. Thirunavukkarasu K, Pei Y, Moore TL, Wang H, Yu XP, Geiser AG,Chandrasekhar S: Regulation of the human ADAMTS-4 promoter bytranscription factors and cytokines. Biochem Biophys Res Commun 2006,345:197–204.

39. Beck GR Jr, Zerler B, Moran E: Gene array analysis of osteoblastdifferentiation. Cell Growth Differ 2001, 12:61–83.

40. Baldridge D, Shchelochkov O, Kelley B, Lee B: Signaling pathways inhuman skeletal dysplasias. Annu Rev Genomics Hum Genet 2010,11:189–217.

41. Lee BK, Bhinge AA, Battenhouse A, McDaniell RM, Liu Z, Song L, Ni Y, BirneyE, Lieb JD, Furey TS, Crawford GE, Iyer VR: Cell-type specific andcombinatorial usage of diverse transcription factors revealed bygenome-wide binding studies in multiple human cells. Genome Res 2012,22:9–24.

42. Birney E, Stamatoyannopoulos J, Dutta A, Guigó R, Gingeras TR, MarguliesEH, Weng Z, Snyder M, Dermitzakis ET, Thurman RE, Kuehn MS, Taylor CM,Neph S, Koch CM, Asthana S, Malhotra A, Adzhubei I, Greenbaum J,Andrews RM, Flicek P, Boyle PJ, Cao H, Carter NP, Clelland GK, Davis S, DayN, Dhami P, Dillon SC, Dorschner MO, Fiegler H, et al: Identification andanalysis of functional elements in 1% of the human genome by theENCODE pilot project. Nature 2007, 447:799–816.

43. Neph S, Vierstra J, Stergachis AB, Reynolds AP, Haugen E, Vernot B, ThurmanRE, John S, Sandstrom R, Johnson AK, Maurano MT, Humbert R, Rynes E,Wang H, Vong S, Lee K, Bates D, Diegel M, Roach V, Dunn D, Neri J, SchaferA, Hansen RS, Kutyavin T, Giste E, Weaver M, Canfield T, Sabo P, Zhang M,Balasundaram G, et al: An expansive human regulatory lexicon encodedin transcription factor footprints. Nature 2012, 489:83–90.

44. Farnham PJ: Insights from genomic profiling of transcription factors.Nat Rev Genet 2009, 10:605–616.

45. Robertson G, Hirst M, Bainbridge M, Bilenky M, Zhao Y, Zeng T, EuskirchenG, Bernier B, Varhol R, Delaney A, Thiessen N, Griffith OL, He A, Marra M,Snyder M, Jones S: Genome-wide profiles of STAT1 DNA associationusing chromatin immunoprecipitation and massively parallelsequencing. Nat Methods 2007, 4:651–657.

46. Stender JD, Kim K, Charn TH, Komm B, Chang KC, Kraus WL, Benner C,Glass CK, Katzenellenbogen BS: Genome-wide analysis of estrogenreceptor alpha DNA binding and tethering mechanisms identifies Runx1as a novel tethering factor in receptor-mediated transcriptional activation.Mol Cell Biol 2010, 30:3943–3955.

47. Tijssen MR, Cvejic A, Joshi A, Hannah RL, Ferreira R, Forrai A, Bellissimo DC,Oram SH, Smethurst PA, Wilson NK, Wang X, Ottersbach K, Stemple DL,Green AR, Ouwehand WH, Gottgens B: Genome-wide analysis ofsimultaneous GATA1/2, RUNX1, FLI1, and SCL binding in megakaryocytesidentifies hematopoietic regulators. Dev Cell 2011, 20:597–609.

48. Visel A, Blow MJ, Li Z, Zhang T, Akiyama JA, Holt A, Plajzer-Frick I, ShoukryM, Wright C, Chen F, Afzal V, Ren B, Rubin EM, Pennacchio LA: ChIP-seqaccurately predicts tissue-specific activity of enhancers. Nature 2009,457:854–858.

49. Weltmeier F, Borlak J: A high resolution genome-wide scan of HNF4alpharecognition sites infers a regulatory gene network in colon cancer.PLoS One 2011, 6:e21667.

50. Little GH, Noushmehr H, Baniwal SK, Berman BP, Coetzee GA, Frenkel B:Genome-wide Runx2 occupancy in prostate cancer cells suggests a rolein regulating secretion. Nucleic Acids Res 2012, 40:3538–3547.

51. Javed A, Barnes GL, Jasanya BO, Stein JL, Gerstenfeld L, Lian JB: Stein GS:runt homology domain transcription factors (Runx, Cbfa, and AML)mediate repression of the bone sialoprotein promoter: evidence forpromoter context-dependent activity of Cbfa proteins. Mol Cell Biol 2001,21:2891–2905.

52. Paredes R, Arriagada G, Cruzat F, Villagra A, Olate J, Zaidi K, van Wijnen A,Lian JB, Stein GS, Stein JL, Montecino M: Bone-specific transcription factorRunx2 interacts with the 1alpha,25-dihydroxyvitamin D3 receptor toup-regulate rat osteocalcin gene expression in osteoblastic cells.Mol Cell Biol 2004, 24:8847–8861.

53. Paz J, Wade K, Kiyoshima T, Sodek J, Tang J, Tu Q, Yamauchi M, Chen J:Tissue- and bone cell-specific expression of bone sialoprotein isdirected by a 9.0 kb promoter in transgenic mice. Matrix Biol 2005,24:341–352.

54. Roca H, Phimphilai M, Gopalakrishnan R, Xiao G, Franceschi RT: Cooperativeinteractions between RUNX2 and homeodomain protein-binding sitesare critical for the osteoblast-specific expression of the bone sialoproteingene. J Biol Chem 2005, 280:30845–30855.

55. Li G, Ruan X, Auerbach RK, Sandhu KS, Zheng M, Wang P, Poh HM, Goh Y,Lim J, Zhang J, Sim HS, Peh SQ, Mulawadi FH, Ong CT, Orlov YL, Hong S,Zhang Z, Landt S, Raha D, Euskirchen G, Wei CL, Ge W, Wang H, Davis C,Fisher-Aylor KI, Mortazavi A, Gerstein M, Gingeras T, Wold B, Sun Y, et al:Extensive promoter-centered chromatin interactions provide atopological basis for transcription regulation. Cell 2012, 148:84–98.

56. Kowalczyk MS, Hughes JR, Garrick D, Lynch MD, Sharpe JA, Sloane-StanleyJA, McGowan SJ, De Gobbi M, Hosseini M, Vernimmen D, Brown JM, GrayNE, Collavin L, Gibbons RJ, Flint J, Taylor S, Buckle VJ, Milne TA, Wood WG,Higgs DR: Intragenic enhancers act as alternative promoters. Mol Cell2012, 45:447–458.

57. Lieberman-Aiden E, van Berkum NL, Williams L, Imakaev M, Ragoczy T,Telling A, Amit I, Lajoie BR, Sabo PJ, Dorschner MO, Sandstrom R, BernsteinB, Bender MA, Groudine M, Gnirke A, Stamatoyannopoulos J, Mirny LA,Lander ES, Dekker J: Comprehensive mapping of long-range interactionsreveals folding principles of the human genome. Science 2009,326:289–293.

58. Tan-Wong SM, Zaugg JB, Camblong J, Xu Z, Zhang DW, Mischo HE, Ansari AZ,Luscombe NM, Steinmetz LM, Proudfoot NJ: Gene loops enhancetranscriptional directionality. Science 2012, 338:671–675.

59. Zaidi SK, Javed A, Choi JY, van Wijnen AJ, Stein JL, Lian JB, Stein GS:A specific targeting signal directs Runx2/Cbfa1 to subnuclear domainsand contributes to transactivation of the osteocalcin gene. J Cell Sci 2001,114:3093–3102.

60. Young DW, Hassan MQ, Pratap J, Galindo M, Zaidi SK, Lee SH, Yang X, Xie R,Javed A, Underwood JM, Furcinitti P, Imbalzano AN, Penman S, NickersonJA, Montecino MA, Lian JB, Stein JL, van Wijnen AJ, Stein GS: Mitoticoccupancy and lineage-specific transcriptional control of rRNA genes byRunx2. Nature 2007, 445:442–446.

61. FoxPrimer qPCR Primer Design Suite. [http://www.foxprimer.org]62. Dobson JR: Nuclear organization in breast cance. In PhD thesis.: UMASS

Medical School; 2013.63. Pratap J, Wixted JJ, Gaur T, Zaidi SK, Dobson J, Gokul KD, Hussain S,

van Wijnen AJ, Stein JL, Stein GS, Lian JB: Runx2 transcriptional activationof Indian Hedgehog and a downstream bone metastatic pathway inbreast cancer cells. Cancer Res 2008, 68:7795–7802.

64. R: A Language and Environment for Statistical Computing. [http://www.R-project.org]

http://www.foxprimer.org

http://www.r-project.org

http://www.r-project.org


65. Gautier L, Cope L, Bolstad BM, Irizarry RA: affy–analysis of AffymetrixGeneChip data at the probe level. Bioinformatics 2004, 20:307–315.

66. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B,Gautier L, Ge Y, Gentry J, Hornik K, Hothorn T, Huber W, Iacus S, Irizarry R,Leisch F, Li C, Maechler M, Rossini AJ, Sawitzki G, Smith C, Smyth G, TierneyL, Yang JY, Zhang J: Bioconductor: open software development forcomputational biology and bioinformatics. Genome Biol 2004, 5:R80.

67. Smyth GK: Limma: linear models for microarray data. In Bioinformatics andComputational Biology Solutions Using R and Bioconductor. Edited byGentleman R, Carey VJ, Huber W, Irizarry RA, Dudoit S. New York: Springer;2005:397–420 [Gail M, Krickeberg K, Samet J, Tsiatis A, Wong W (SeriesEditors): Statistics for Biology and Health].

68. Benjamini Y, Hochberg Y: Controlling the false discovery rate - a practicaland powerful approach to multiple testing. J R Stat Soc Series B Methodol1995, 57:289–300.

69. Dignam JD, Lebovitz RM, Roeder RG: Accurate transcription initiation byRNA polymerase II in a soluble extract from isolated mammalian nuclei.Nucleic Acids Res 1983, 11:1475–1489.

70. van der Deen M, Akech J, Lapointe D, Gupta S, Young DW, Montecino MA,Galindo M, Lian JB, Stein JL, Stein GS, van Wijnen AJ: Genomic promoteroccupancy of runt-related transcription factor RUNX2 in Osteosarcomacells identifies genes involved in cell adhesion and motility. J Biol Chem2012, 287:4503–4517.

71. Langmead B, Trapnell C, Pop M, Salzberg SL: Ultrafast and memory-efficient alignment of short DNA sequences to the human genome.Genome Biol 2009, 10:R25.

72. Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM, HausslerD: The human genome browser at UCSC. Genome Res 2002, 12:996–1006.

73. McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB, Wenger AM,Bejerano G: GREAT improves functional interpretation of cis-regulatoryregions. Nat Biotechnol 2010, 28:495–501.

doi:10.1186/gb-2014-15-3-r52Cite this article as: Wu et al.: Genomic occupancy of Runx2 with globalexpression profiling identifies a novel dimension to control ofosteoblastogenesis. Genome Biology 2014 15:R52.

Submit your next manuscript to BioMed Centraland take full advantage of:

• Convenient online submission

• Thorough peer review

• No space constraints or color figure charges

• Immediate publication on acceptance

• Inclusion in PubMed, CAS, Scopus and Google Scholar

• Research which is freely available for redistribution

Submit your manuscript at www.biomedcentral.com/submit

research open access genomic occupancy of runx2 with ... · research open access genomic occupancy...

Documents